This repository has been archived by the owner on Oct 19, 2024. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 360
Issues: alpa-projects/alpa
Author
Label
Projects
Milestones
Assignee
Sort
Issues list
[FEATURE] Reduce the required peak RAM on a single node while converting weights
good first issue
Good for newcomers
#792
opened Dec 1, 2022 by
zhanyuanucb
[PERF] Port the manually optimized CUDA kernels into Alpa
enhancement
New feature
good first issue
Good for newcomers
#406
opened Apr 27, 2022 by
zhisbug
[RUNTIME] TPU Backend Support
enhancement
New feature
#433
opened May 7, 2022 by
merrymercy
1 of 2 tasks
[PERF] Save/Restore RNG States for Rematerialization
enhancement
New feature
#517
opened Jun 16, 2022 by
merrymercy
[Roadmap] Inference performance (for OPT/GPT)
enhancement
New feature
#534
opened Jun 21, 2022 by
zhisbug
2 of 8 tasks
[BUG] Balance assignment of free equations in PipeshardParallel
good first issue
Good for newcomers
known bug
Something isn't working
#549
opened Jun 27, 2022 by
merrymercy
[FEATURE] Auto-Parallelization for Graphs with Predesignated Parameter Shardings
enhancement
New feature
#550
opened Jun 28, 2022 by
zhuohan123
[BUG] PipeshardParallel crashes when apply_grad part is empty
known bug
Something isn't working
#560
opened Jun 30, 2022 by
merrymercy
[BUG] rng runs with each microbatch
known bug
Something isn't working
#592
opened Jul 6, 2022 by
ZYHowell
[BUG] Inaccurate memory profiling during pipeline stage construction
known bug
Something isn't working
#684
opened Aug 31, 2022 by
zhuohan123
[FEATURE] A CPU Swapping Runtime
enhancement
New feature
good first issue
Good for newcomers
#694
opened Sep 9, 2022 by
merrymercy
3 tasks
IndexError:
InlinedVector::at(size_type) const
failed bounds check
unknown error
#786
opened Nov 27, 2022 by
TonyTangYu
[FEATURE] Current gradient accumulation only works for jnp.mean loss
enhancement
New feature
#268
opened Dec 16, 2021 by
zhuohan123
[BUG] Collective group's rank is incorrect
good first issue
Good for newcomers
known bug
Something isn't working
#790
opened Dec 1, 2022 by
ZYHowell
2 tasks
[FEATURE] launch physical meshes after compilation
enhancement
New feature
good first issue
Good for newcomers
#801
opened Dec 6, 2022 by
ZYHowell
3 tasks
[FEATURE] Reduce congestion of sending on one mesh
enhancement
New feature
good first issue
Good for newcomers
#802
opened Dec 6, 2022 by
ZYHowell
4 tasks
'a:f32[32,12,512,512] = add_any b c' cannot be delayed in apply_grad.py
unknown error
#808
opened Dec 7, 2022 by
cksmll
soft lockup - CPU#113 stuck after terminating training process
unknown error
#814
opened Dec 16, 2022 by
frankxyy
Training throughput not scale with increasing number of devices (ViT training)
#824
opened Dec 20, 2022 by
frankxyy
[PERF] Bind constant tensors for an executable (i.e. bind parameters for inference)
good first issue
Good for newcomers
#843
opened Jan 3, 2023 by
merrymercy
2 tasks
Dreambooth fine-tune support
help wanted
Extra attention is needed
#851
opened Jan 11, 2023 by
huanchengying
Previous Next
ProTip!
What’s not been updated in a month: updated:<2024-12-10.